人工鱼群算法在SVM参数优化选择中的应用

Application of artificial fish-swarm algorithm in SVM parameter optimization selection

针对支持向量机的参数优化缺乏理论支持,而SVM交叉检验法选取又较为费时的情况下,提出了基于人工鱼群算法的支持向量机参数优化选取算法,并以SVM分类预测准确率最大为优化原则,利用人工鱼群算法的较好并行性和较强的全局寻优能力,以实现最优目标并得到SVM的最优参数组合。数值实验结果表明:人工鱼群算法在SVM参数优化选取中具有更快的寻优性能,同时具有较高的分类准确率。该方法具有较好的并行性和较强的全局寻优能力。

As considering that the parameter optimization of support vector machine lacks theory support and the SVM cross-validation method spends lots of time on selecting parameters, the parameter optimization selection method of support vec- tor machine is proposed based on artificial fish-swarm algorithm. This method puts the SVM classification prediction accuracy rate as the optimization principle and uses the better parallelism of artificial fish-swarm algorithm and the stronger global optimi- zation ability to achieve the optimal target and obtain optimal parameter combination of SVM. The results of numerical value experi- ments show that the artificial fish-swarm algorithm has faster performance optimization and higher classification accuracy rate in SVM parameters＇ optimization selection. This method has the better parallelism and the stronger global optimization ability.